Benefits and challenges of using tensor-network quantum emulators for Hamiltonian simulation
One of the most promising applications of quantum computing is high accuracy simulation of quantum chemistry systems, required for the precise modelling of drugs. Despite the tremendous progress in quantum hardware development and error correction protocols, current day hardware is still susceptible to errors. To overcome this limitation, quantum emulators can be used instead to run quantum algorithms at an intermediate scale on conventional computers, serving as a bridge to fault-tolerant quantum computing.
Making use of Fermioniq's quantum emulator Ava, Capgemini's Quantum Lab investigated the benefits and challenges of using tensor-network emulators within a machine-learning pipeline for the prediction of reactivity for target covalent drugs. Their findings can be found in this Medium post. Using Ava makes it possible to simulate larger quantum systems than conventional statevector methods can, allowing for an increase in the active space size of the molecules studied.
The role of the quantum computer, or emulator in this case, in the machine-learning pipeline is to compute energies of a time-evolved quantum state, which serve as a 'quantum fingerprint', from which the reactivity of a target drug can be estimated. Time-evolution is one of the most challenging tasks for quantum emulators, especially when emulating larger quantum systems for which the statevector is too large to fit in memory, and more sophisticated methods -- such as the tensor network methods employed by Ava -- are required. Typically, as time evolves, entanglement gets built up in the quantum state, and as the amount of entanglement increases, more resources are needed for the tensor network method to accurately capture the quantum state in question. The amount of resources used for a tensor network simulation is given by a single parameter, called the bond dimension.
The team at Capgemini that performed the research came to the conclusion that, as the active space size (given by the number of qubits) increases, the bond dimension required to simulate the molecule in question within chemical accuracy increases rapidly; making a strong case for fault-tolerant quantum hardware which, being quantum itself, naturally captures the entanglement present in the quantum state of interest.